[amsat-bb] Re: Decoding wideband recordings

[Gordon JC Pearce]

On Tue, 09 Aug 2011 15:54:15 -0700
Phil Karn <karn at philkarn.net> wrote:

> It's been suggested that I modify my ARISSat-1 BPSK-1000 telemetry
> demodulator/decoder to accept wideband quadrature (I & Q) recordings
> like those produced by most of the software defined radios out there.
> 
> This is fundamentally not that hard, but first I need some information.
> 
> How many people could actually use this? What is the format of the
> recorded files? Is there a standard, or does each make of SDR produce
> its own?

Anyone with a Funcube Dongle, for a start.
 
> If I could get a good sample wideband quadrature recording of the
> ARISSat-1 downlink that includes the BPSK beacon, I could use it for
> testing. Ideally there would be no need for any manual tuning because
> the wideband recording would always contain the beacon regardless of
> Doppler. All tuning would be in the demodulator software.

If you do get one, I'd like a copy as well.  The tuning bit is easy, you just multiply the incoming complex signal by a rotating vector.  Want to tune higher?  Spin that vector a little faster!

> This might be a little compute-intensive so I can't guarantee that it'll
> run in real time on older computers. I already make pretty heavy use of
> the Intel SSE (vector arithmetic) instructions for FIR filtering and
> Viterbi decoding. The SSE2 set introduced with the Pentium 4 has been
> especially useful, so I would consider the P4 (or the AMD equivalent) a
> minimum requirement. Of course, the newer and faster the CPU, the
> better. I also make use of multiple CPU cores, particularly during the
> acquisition phase, so the newer CPUs with lots of CPU cores would also
> be preferred.

I've successfully run lysdr on a 1400MHz Celeron M at 48kHz, although I had to back off the screen updates.  The FFT is actually quite fast, but screen updates with the Intel chipset tend to slow things down.  I considered using FFT convolution to replace the FIR filter, because the FIR is *really* slow.

https://github.com/gordonjcp/lysdr



-- 
Gordon JC Pearce MM0YEQ <gordonjcp at gjcp.net>